try fixing GPU (torch)

requirements.txt

@@ -38,16 +38,25 @@ duckdb>=0.9.0
 Flask>=3.0.0
 flask-cors>=4.0.0
 
-# NOTE: PyTorch is intentionally not pinned here to avoid pulling NVIDIA CUDA deps on AMD machines.
-# Install one of the following sets manually depending on your hardware:
+# NOTE: PyTorch is intentionally not pinned here to avoid pulling wrong GPU deps.
+# We only need torch (not torchvision/torchaudio) for trading systems.
 #
-# CPU-only (AMD/Intel, no NVIDIA CUDA):
-#   pip install torch torchvision torchaudio --index-url https://download.pytorch.org/whl/cpu
+# The SAME CODEBASE works with NVIDIA (CUDA) and AMD (ROCm) GPUs!
+# PyTorch abstracts hardware differences - just install the right build.
 #
-# NVIDIA GPU (CUDA):
-#   Visit https://pytorch.org/get-started/locally/ for the correct command for your CUDA version.
-#   Example (CUDA 12.1):
-#   pip install torch torchvision torchaudio --index-url https://download.pytorch.org/whl/cu121
+# AUTOMATIC SETUP (recommended):
+#   ./scripts/setup-pytorch.sh
 #
-# AMD Strix Halo NPU Acceleration:
-#   pip install onnxruntime-directml onnx transformers optimum
+# MANUAL INSTALL by hardware:
+#
+# CPU-only (development/testing):
+#   pip install torch --index-url https://download.pytorch.org/whl/cpu
+#
+# NVIDIA GPU (CUDA 12.1):
+#   pip install torch --index-url https://download.pytorch.org/whl/cu121
+#
+# AMD GPU (ROCm 6.2):
+#   pip install torch --index-url https://download.pytorch.org/whl/rocm6.2
+#
+# Verification:
+#   python -c "import torch; print(f'GPU: {torch.cuda.is_available()}')"